Assigning Human-Understandable Labels to Web Pages

• US 20090319533A1
• Filed: 06/23/2008
• Published: 12/24/2009
• Est. Priority Date: 06/23/2008
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Tengli, Ashwin
• Granted Patent: US 8,185,528 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/95 Retrieval from the web

Abstract

Methods and systems that label a web page collect a set of inbound labels for the web page, estimate a language model for the web page, compute the likelihood of generating each inbound label given the language model and assign a score to each inbound label based on this likelihood, and assign a label to the web page based on the score assigned to each of the set of inbound labels. Inbound labels are preferably collected from the set of web documents linking to the web page. Labels assigned are useful in providing labeled links to web pages from top hosts in search result pages.

Claims

1. A computer-implemented method of labeling a web page from a host, comprising:

a. estimating a language model for the web page;

b. collecting, from the set of web documents linking to the web page, a set of inbound labels for the web page;

c. computing the likelihood of generating each inbound label given the language model and assigning a score to each inbound label based on this likelihood; and

d. assigning a label to the web page based on the scores assigned to each of the set of inbound labels.

2. The computer-implemented method of claim 1, wherein a set of language models are estimated for each web page.

3. The computer-implemented method of claim 2, wherein computing the likelihood proceeds via interpolation between the language models in the set of language models.

4. The computer-implemented method of claim 2, wherein computing the likelihood proceeds via forming a meta-classifier based on the set of language models.

5. The computer-implemented method of claim 2, wherein a language model estimated for the web page incorporates the set of inbound labels for the web page.

6. The computer-implemented method of claim 1, wherein the set of inbound labels includes text obtained from anchor text in links to the web page.

7. The computer-implemented method of claim 1, wherein the set of inbound labels includes text obtained from tags associated with the web page on social-tagging sites.

8. The computer-implemented method of claim 1, wherein the set of inbound labels includes text obtained from search queries that have returned the web page.

9. An offline processing module for labeling a web page from a host, comprising:

a. a label collection element configured to collect, from the set of web documents linking to the web page, a set of inbound labels for the web page;

b. a language model estimator configured to estimate a language model for the web page;

c. a computation element configured to compute the likelihood of generating each inbound label given the language model and to assign a score to each inbound label based on this likelihood; and

d. a label assignment element configured to assign a label to the web page from the set of inbound labels based on the scores assigned to each inbound label.

10. The offline processing module of claim 9, wherein a set of language models are estimated for each web page.

11. The offline processing module of claim 10, wherein computing the likelihood proceeds via interpolation between the language models in the set of language models.

12. The offline processing module of claim 10, wherein computing the likelihood proceeds via forming a meta-classifier based on the set of language models.

13. The offline processing module of claim 10, wherein a language model estimated for the web page incorporates the set of inbound labels for the web page.

14. The offline processing module of claim 9, wherein the set of inbound labels includes text obtained from anchor text in links to the web page.

15. The offline processing module of claim 9, wherein the set of inbound labels includes text obtained from tags associated with the web page on social-tagging sites.

16. The offline processing module of claim 9, wherein the set of inbound labels includes text obtained from search queries that have returned the web page.

17. A system for providing labeled links to top web pages from a top host in a search results page, comprising:

a. an offline processing module configured to provide labels for top web pages from top hosts by collecting a set of inbound labels for the web pages, estimating a language model for each web page, and assigning label to each web page based on a computation involving the language model;

b. an element configured to select a set of top web pages from a top host; and

c. an online module that publishes the labels of each of the top web pages of the top host in the search result for the top host.

18. The system of claim 17, wherein the computation finds likelihood of generating each inbound label given the language model and assigns a score to each inbound label based on this likelihood.

19. The system of claim 18, wherein a set of language models are estimated for each web page.

20. The system of claim 19, wherein computing the likelihood proceeds via interpolation between the language models in the set of language models.

21. The system of claim 19, wherein computing the likelihood proceeds via forming a meta-classifier based on the set of language models.

22. The system of claim 19, wherein a language model estimated for the web page incorporates the set of inbound labels for the web page.

23. The system of claim 17, wherein the set of inbound labels includes text obtained from anchor text in links to the web page.

24. The system of claim 17, wherein the set of inbound labels includes text obtained from tags associated with the web page on social-tagging sites.

25. The system of claim 17, wherein the set of inbound labels includes text obtained from search queries that have returned the web page.